On October 5, OpenAI announced a new visual ad format for ChatGPT, built specifically around image generation sessions. The ads are designed to appear alongside images that people create inside ChatGPT, and OpenAI says it will start testing the format in the United States with a limited set of advertisers. For Korean and Japanese consumer brands selling in the U.S., the honest answer to “should we jump in” is: not yet with real budget, but yes with a small, well-measured pilot once access opens. This piece walks through what was actually announced, why an image session is a different animal from a search result, and the handful of things worth testing first.
Key takeaways (30-second version)
- It is a test, not a launch: OpenAI is starting with a limited group of advertisers in the U.S., so most brands will wait for access before spending.
- The surface is visual, not textual: ads appear next to images people generate, which rewards product look, packaging, and mood over keyword copy.
- Measurement came with it: the announcement paired the format with expanded ad-measurement tools and attribution partnerships, which signals OpenAI knows advertisers need proof.
- Asian consumer brands have a visual edge: K-beauty textures, Japanese snack packaging, and home goods photograph well, which suits an image-first placement.
- Start small and instrument everything: treat the first run as a learning budget with clean tracking, not a scaling play.
1. What OpenAI actually announced
The announcement is narrow and specific, which is a good thing for anyone trying to plan around it. OpenAI introduced a visual ad format for ChatGPT that centers on image generation sessions. The ads are meant to show up alongside the images people produce when they ask ChatGPT to create or edit a picture. OpenAI said it plans to begin testing this in the United States with a limited set of advertisers, rather than opening it to everyone at once.
The second part of the news matters as much as the format itself. According to the announcement, the rollout includes expanded ad-measurement tools and partnerships with attribution providers. In plain terms, OpenAI is not just placing a picture next to a picture. It is trying to give advertisers a way to connect an ad view to an outcome. That pairing of format plus measurement is the detail worth circling, because a new placement without measurement is a guess, and a new placement with measurement is a test you can actually read.
A few things reported around the announcement have not been corroborated by independent sources, so treat them as open questions rather than facts. That includes any specific start date for when ads begin appearing, the exact list of attribution partners, and how this compares to other companies piloting ads in AI assistants. We will not state those as settled here.
2. Why an image session is not a search box
Most performance marketers built their instincts on search. Someone types a query, a result or ad answers it, and intent is fairly legible in the words. An image generation session works on a different logic. The person is not describing a thing they want to buy. They are describing a thing they want to see. The signal is aesthetic and contextual: a mood, a color story, a style of room, a type of face care routine, a snack spread for a party.
That shifts what a good ad looks like. In a visual session, the creative is the pitch. Packaging design, texture, lighting, and how a product sits in a scene carry the weight that headlines carry in search. A brand with a strong visual identity has a natural advantage, and a brand with weak or inconsistent imagery will struggle to earn attention no matter how sharp the copy is.
Why this matters: in an image-first placement, your product photography is your ad budget’s biggest multiplier. A brand that ships beautiful, consistent visuals will get more out of the same spend than one relying on words to do the persuading.
3. Why this suits Korean and Japanese consumer brands
Korean and Japanese consumer categories tend to be visually rich in ways that fit an image session well. K-beauty leans on texture shots, before-and-after skin states, and clean minimalist packaging. Japanese snacks and confectionery carry distinctive package art and playful product forms. Home, stationery, and kitchen goods from both markets often win on design detail that reads instantly in a picture.
That is a real edge on a surface where the image is the message. A U.S. shopper asking ChatGPT to help visualize a skincare shelf, a gift box, or a desk setup is in a receptive, open frame of mind, and a well-styled product that matches the scene can feel like a helpful suggestion rather than an interruption. The brands best positioned here are the ones whose look already does the selling.
The flip side is honesty about where visual placement does not fit. Products that sell on spec sheets, certifications, or detailed claims will find an image slot a poor place to carry that information. For those, this format is an awareness and consideration play at best, and the real conversion work still happens on the product page.
4. What to test first: a practical shortlist
Once access opens, resist the urge to port your whole catalog in. A visual ad test rewards focus. Here is a shortlist worth running, ordered by how cleanly you can read the result.
| Test priority | What to run | Why it is a good first test |
|---|---|---|
| 1. Hero product in a scene | Your single best-selling SKU, styled in a lifestyle context (vanity, kitchen, desk) | Highest-fidelity read on whether the surface drives interest for a proven winner |
| 2. Packaging-forward shot | A clean, design-led product image where the package itself is the draw | Isolates the visual-identity advantage many Asian brands already have |
| 3. Gift or set bundle | A curated set that photographs as a complete idea | Image sessions often involve “help me picture a gift,” which maps to bundles |
| 4. Seasonal or occasion look | A time-bound styling (holiday, back-to-school) tied to a short window | Lets you test relevance of context matching without long commitment |
Across all of these, keep three disciplines. First, hold your landing experience constant so the ad is the only variable. Second, cap the budget at a level you are comfortable treating as tuition, because early tests buy learning, not scale. Third, prepare two or three visual directions per product rather than one, since the whole point of this surface is to learn which look resonates.
Set the bar before you start
Decide your success metric in advance. For a first visual test, a realistic goal is a clean signal on engagement and downstream site behavior, not an immediate return-on-ad-spend target that you would hold a mature search campaign to. Writing the goal down before launch keeps you from reading noise as either triumph or failure.
5. Measurement and attribution readiness
The announcement’s inclusion of expanded measurement tools and attribution partnerships is the part that should shape how you prepare. A new ad surface is only as useful as your ability to see what it did. Before you request access, get your own house in order so that whatever measurement OpenAI provides lands on a clean foundation.
Practically, that means confirming your site analytics capture referral traffic correctly, your tagging is consistent, and you have a baseline for the SKUs you plan to promote. If you cannot describe how a given product performs today, you will not be able to tell whether the ad changed anything tomorrow. Attribution from any new channel is easier to trust when your first-party data is tidy and your conversion events are defined the same way everywhere.
Why this matters: the brands that learn fastest from new ad surfaces are not the ones with the biggest budgets. They are the ones whose measurement was clean enough to read a small test honestly.
6. What we do not know yet
A responsible plan names its blind spots. Several details that would normally drive a media plan are not confirmed. We do not have a verified public start date for when ads begin appearing to users. We do not have an independently confirmed list of the attribution providers involved. We do not have reliable pricing, auction mechanics, or targeting controls. And we do not have a corroborated picture of how this approach differs from other AI-assistant ad pilots running elsewhere.
That uncertainty is a reason to prepare, not to panic. The right posture is to get creative assets and measurement ready now, request access when it is available, and let the first real test teach you the mechanics that the announcement left open. Anyone promising precise performance numbers for a format still in limited testing is guessing, and guessing is an expensive way to plan a budget.
7. Frequently asked questions
Q1. Can any brand buy ChatGPT visual ads right now?
No. OpenAI said it plans to start testing the format in the United States with a limited set of advertisers. Broad availability has not been announced, so most brands should prepare rather than expect immediate access.
Q2. Where exactly do these ads appear?
According to the announcement, the ads are designed to appear alongside images that people generate within ChatGPT, during image generation sessions specifically. This is a visual placement, not a text ad in a chat reply.
Q3. Is this a good fit for K-beauty and Japanese snack brands?
It can be, because these categories are visually distinctive and photograph well, which suits an image-first surface. It is less suited to products that rely on detailed specs or claims that a single image cannot carry.
Q4. How should we measure results?
The rollout reportedly includes expanded ad-measurement tools and attribution partnerships. Pair whatever OpenAI provides with clean first-party analytics and a pre-test baseline so you can read a small pilot honestly.
Q5. Should we move budget out of search or social to fund a test?
Not at first. Treat an early visual ad pilot as a small learning budget funded separately, so a format still in limited testing does not disrupt channels you already understand.
Q6. When will the ads actually go live for users?
A specific public start date has not been independently confirmed. Prepare your assets and measurement now, and let the timing follow access rather than planning around an unverified date.
8. The bottom line
OpenAI’s visual ad format is an early, limited test of a genuinely new surface: ads that live next to the images people create inside ChatGPT. For Korean and Japanese consumer brands, the opportunity is real precisely because the surface rewards strong visual identity, and that is something many of these brands already have. The discipline is to treat the first run as a measured pilot, lead with your best product imagery, set a success bar in advance, and keep your measurement clean enough to learn something true from a small spend.
At Calywire, a U.S.-based marketing agency working with Korean and Japanese consumer brands since 2014, our view is simple: get the creative and the measurement ready now, request access when it opens, and let a small honest test teach you the rest. If you want a second set of eyes on which SKUs to pilot and how to instrument the results, that is a conversation worth having before the window gets crowded.
Sources
